Models to Determine Relationship of Motor Skills on Social Deficit
Nicholas Mobley
12/4/2019
Introduction: Study Details & Relevant Background Information
Autism (ASD) and Attention Deficit/Hyperactivity Disorder (ADHD) are frequently observed in children, but are still not fully understood.
- Researchers at Kennedy Krieger Institute(KKI) ran a study on children with ASD, ADHD, and typically developing children to study differences in social deficits, motor impairment, intelligence.
- It is known that children with developmental disorders suffer from motor impairment concurrent with social deficits associated with their respective diagnoses(1,2). The extent and cause of this correlation merits further research.
- Metrics and Scales Used to Quantify Symptom Severity in Children:
- Social Responsiveness Scale Questionnaire (SRS): measures severity of social deficits in children (higher scores indicate more social deficits)
- Movement Assessment Battery for Children (mABC): measures motor impairments in children. Three subcategories for assessment include dexterity, aiming and catching, and balance summed to obtain the mABC Total Standard Score
- Wechsler Intelligence Scale for Children (WISC): measures intelligence in children. Two versions of the test quantify vocabulary, information, picture concepts, etc.
Introduction: Areas of Particular Interest
Data and Analysis
We want to investigate the overall association between social deficits and motor skills for each diagnosis group.
> Primary Endpoint: What motor skills are associated with SRS scores for each diagnosis group. > Secondary Endpoint: What factors influence this association? - Data Collected for Investigation and Analysis:
- Age
- Handedness (quantified by Edinburgh Handedness Integer)
- Gender
- Primary and Secondary Diagnosis
- SRS Total
- mABC Total and Subscores
- WISC Score
Methods: Cleaning Data
- Removed Subjects that were under 14, since 94% subjects diagnosed with Autism or ADHD were under 14. Reduced total observations from 832 to 755
- Then Removed Observations that had N/A values for either SRS_TotalRawScore or mABC_TotalStandardScore which furthur reduced the total observations to 568 Observations
- Converted all the WISC scores to the GAI scale for comparability
| Primary | Count | Avg Age | Males | Females | RightHanded | LeftHanded |
|---|---|---|---|---|---|---|
| ASD | 138 | 10.236 | 116 | 22 | 124 | 20 |
| ADHD | 176 | 9.6908 | 124 | 52 | 285 | 36 |
| TD | 254 | 10.2146 | 185 | 69 | 307 | 42 |
Methods: Exploring Motor Skills
Methods: Other Variables
- Seems to be negative correlation between Intelligence and Motor Skills, but no correlation between Intelligence and Social Deficit
Proposed Models
\[ \text{E[Social Deficit Score]} = \textbf{X} * \boldsymbol{\beta} \\ where: \\ \boldsymbol{\beta} = \begin{bmatrix} \text{Intercept} \\ \text{Balance Score} \\ \text{Aiming And Catching Score} \\ \text{EdinburgHandedness_Integer} \\ \text{GAI} \\ \text{Gender Male} \\ \text{mABC_AGE} \end{bmatrix} \]
- seperate SRS Versions from Welch t-test
\[ H_0: \text{Mean of SRS Total Scores from Version 1 = Mean of SRS Total Scores from Version 2} \\ H_1: \text{Mean of SRS Total Scores from Version 1} \space \neq \text{Mean of SRS Total Scores from Version 2} \\ \text{Under the null we got a t-statistc of -6.03 from a t-distribution of 417.38 Degrees of Freedom} \\ \text{that has a p-value of} \space 3.511 \times 10^{-9} \\ \space \text{Under} \space \alpha = .05 \space \text{we reject the null} \text{ and believe that there is a true difference in the mean} \]
Resutls: Typically Developing
##
## Call:
## lm(formula = SRS_TotalRawScore ~ mABC_AimingAndCatching.Component.StandardScore +
## mABC_Balance.Component.StandardScore + EdinburghHandedness_Integer +
## GAI + Gender + mABC_AGE, data = nodia2)
##
## Residuals:
## Min 1Q Median 3Q Max
## -6.9465 -2.7041 -0.5293 1.5000 18.0316
##
## Coefficients:
## Estimate Std. Error t value
## (Intercept) 43.52523 4.62720 9.406
## mABC_AimingAndCatching.Component.StandardScore 0.01693 0.19383 0.087
## mABC_Balance.Component.StandardScore -0.14406 0.16893 -0.853
## EdinburghHandedness_Integer -1.46571 1.07791 -1.360
## GAI 0.05956 0.03199 1.862
## GenderM -1.12175 1.24382 -0.902
## mABC_AGE -0.09471 0.42545 -0.223
## Pr(>|t|)
## (Intercept) 1.55e-15 ***
## mABC_AimingAndCatching.Component.StandardScore 0.9306
## mABC_Balance.Component.StandardScore 0.3958
## EdinburghHandedness_Integer 0.1769
## GAI 0.0655 .
## GenderM 0.3692
## mABC_AGE 0.8243
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 4.727 on 103 degrees of freedom
## (1 observation deleted due to missingness)
## Multiple R-squared: 0.06594, Adjusted R-squared: 0.01153
## F-statistic: 1.212 on 6 and 103 DF, p-value: 0.3062Results: ADHD
- For SRS Version 1
\[ \boldsymbol{\beta} = \begin{bmatrix} \text{Intercept} \\ \text{Balance Score} \\ \text{Aiming And Catching Score} \\ \text{EdinburgHandedness_Integer} \\ \text{GAI} \\ \text{ADHD Subtype Inattentive} \\ \text{Gender Male} \\ \text{mABC_AGE} \end{bmatrix} = \begin{bmatrix} 64.0176 \\ -2.3512 \\ 1.3089 \\ 8.3848 \\ 0.0620 \\ -13.9720 \\ -9.3211 \\ -1.1236 \end{bmatrix} \]
## Analysis of Variance Table
##
## Model 1: SRS_TotalRawScore ~ mABC_AimingAndCatching.Component.StandardScore +
## mABC_Balance.Component.StandardScore
## Model 2: SRS_TotalRawScore ~ mABC_AimingAndCatching.Component.StandardScore +
## mABC_Balance.Component.StandardScore + EdinburghHandedness_Integer +
## GAI + ADHD_Subtype + Gender + mABC_AGE
## Res.Df RSS Df Sum of Sq F Pr(>F)
## 1 90 32488
## 2 85 25962 5 6526.4 4.2735 0.001626 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1\[ H_0: \text{Just Balance and Catching is Sufficient} \\ H_1: \text{Full Model is Sufficient} \\ \text{Under the null we got an F-statistic of 4.2735 where} \ F \sim F_{5,85} \\ \text{which gives us a p-value of .0016 with an} \ \alpha = .05 \ \text{we reject the null} \\ \text{and believe that we should keep every coefficient in the full model} \]
- For SRS Version 2
\[ \boldsymbol{\beta} = \begin{bmatrix} \text{Intercept} \\ \text{Balance Score} \\ \text{Aiming And Catching Score} \\ \text{EdinburgHandedness_Integer} \\ \text{GAI} \\ \text{ADHD Subtype Inattentive} \\ \text{Gender Male} \\ \text{mABC_AGE} \end{bmatrix} = \begin{bmatrix} 54.637 \\ 0.2525 \\ -0.9143 \\ 0.4960 \\ 0.0037 \\ -8.7537 \\ 1.6450 \\ 0.5879 \end{bmatrix} \]
## Analysis of Variance Table
##
## Model 1: SRS_TotalRawScore ~ mABC_AimingAndCatching.Component.StandardScore +
## mABC_Balance.Component.StandardScore
## Model 2: SRS_TotalRawScore ~ mABC_AimingAndCatching.Component.StandardScore +
## mABC_Balance.Component.StandardScore + EdinburghHandedness_Integer +
## GAI + ADHD_Subtype + Gender + mABC_AGE
## Res.Df RSS Df Sum of Sq F Pr(>F)
## 1 77 4764.4
## 2 72 4027.4 5 736.99 2.6351 0.03034 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1\[ H_0: \text{Just Balance and Catching is Sufficient} \\ H_1: \text{Full Model is Sufficient} \\ \text{Under the null we got an F-statistic of 2.6351 where} \ F \sim F_{5,72} \\ \text{which gives us a p-value of .03 with an} \ \alpha = .05 \ \text{we reject the null} \\ \text{and believe that we should keep every coefficient in the full model} \]
Results: ASD
- For SRS Version 1
\[ \boldsymbol{\beta} = \begin{bmatrix} \text{Intercept} \\ \text{Balance Score} \\ \text{Aiming And Catching Score} \\ \text{EdinburgHandedness_Integer} \\ \text{GAI} \\ \text{Secondary Diagnosis Yes} \\ \text{Gender Male} \\ \text{mABC_AGE} \end{bmatrix} = \begin{bmatrix} 89.60022 \\ 0.47326 \\ -2.28800 \\ 13.96288 \\ -0.01026 \\ 1.25663 \\ 0.45815 \\ 0.87001 \end{bmatrix} \]
- For SRS Version 2
\[ \boldsymbol{\beta} = \begin{bmatrix} \text{Intercept} \\ \text{Balance Score} \\ \text{Aiming And Catching Score} \\ \text{EdinburgHandedness_Integer} \\ \text{GAI} \\ \text{Secondary Diagnosis Yes} \\ \text{Gender Male} \\ \text{mABC_AGE} \end{bmatrix} = \begin{bmatrix} 58.690482 \\ 0.179097 \\ 0.767967 \\ -2.043116 \\ 0.113743 \\ 8.627154 \\ -4.782354 \\ 0.002163 \end{bmatrix} \]
